CN103679300B - Time forecasting method and device - Google Patents

Abstract

The invention provides a time forecasting method and device and relates to the filed of dynamic traffic information service. The time forecasting method and device can improve accuracy of travel time forecasting on the unusual traffic condition. The time forecasting method includes dividing a path from the departing place to the destination into N links; dividing the N links into X link sets according to the relevance of historical travel time of the N links; acquiring link matrixes corresponding to the X link sets respectively, and making the time in the preset time interval before the forecasting time to be the historical time; performing Autoregressive Integrated Moving Average (ARIMA) training on the link matrixes corresponding to X link sets respectively to obtain the travel time of the X link sets at the forecasting time; acquiring the sum of the travel time of the X link sets at the forecasting time to serve as the travel forecasting time of the forecasting time from the departing place to the destination. The time forecasting method and device are used for travel time forecasting.

Description

A kind of time forecasting methods and device

Technical field

The present invention relates to dynamic information service field, more particularly, to a kind of time forecasting methods and device.

Background technology

Predicting travel time is the important component part of traffic-information service, by predicting travel time, not only can have Effect ground carries out active path planning, thus evading congestion points, reasonably can also plan the departure time, evading the congestion time period.

Road chain hourage historical data by week characteristic day and in one day each time period present periodic feature, and More obvious continuity Characteristics are presented on the time, is therefore based on seasonal effect in time series road chain predicting travel time more universal. Its total thought and method are：The hourage extracting a certain concrete road chain first in characteristic day each time period in each week is equal Then historical travel time and this average are done difference, are obtained a stable time series, finally to this stationary time sequence by value Row autoregression model or mobile autoregression model are modeled and predict.

Because its statistical property of traffic under normal circumstances does not change over time, above method for The prediction of normal traffic is more accurate, and the spies such as accident and large-scale activity in described normal traffic Traffic in the case of different.But, due to being often subject to the impact of event, its statistical property is simultaneously unstable for current urban transportation Fixed, urban traffic conditions assume correlation spatially, and that is, in the event of congestion, then congestion can be along its adjacent road for downstream road chain Chain upstream passing, according to the method described above differentiated time series be difficult to meet stationarity and require, therefore in improper friendship Under logical situation, the time prediction degree of accuracy is relatively low.

Content of the invention

The present invention provides a kind of time forecasting methods and device, can be under improper traffic, when improving travelling Between prediction the degree of accuracy.

For reaching above-mentioned purpose, embodiments of the invention adopt the following technical scheme that：

On the one hand, a kind of time forecasting methods are provided, including：

Path between departure place to destination is divided into N bar link, described N is more than or equal to 2；

Described N bar link is divided into by X link according to the historical travel time correlation degree between described N bar link link Set, described X is less than or equal to described N；

Obtain the corresponding link metric of each link set in described X link set, each described link set pair The link metric answered was made up of the hourage of m historical juncture of all links in described link set, the described historical juncture For the moment in preset time period before prediction time；

Difference auto regressive moving average is carried out to the corresponding link metric of each link set in described X link set Model ARIMA training obtains the hourage in prediction time for the described X link set；

The hourage sum obtaining described X link set in prediction time is as departure place described in described prediction time Travelling predicted time between destination.

On the one hand, a kind of time prediction device is provided, including：

First division unit, for the path between departure place to destination is divided into N bar link, described N is more than or waits In 2；

Second division unit, for according to the historical travel time correlation degree between described N bar link link by described N bar Link is divided into X link set, and described X is less than or equal to described N；

First acquisition unit, for obtaining the corresponding link metric of each link set in described X link set, often The corresponding link metric of individual described link set is by the hourage group of m historical juncture of all links in described link set Become, the described historical juncture is the moment in preset time period before prediction time；

Difference unit, for carrying out difference to the corresponding link metric of each link set in described X link set ARMA model ARIMA training obtains the hourage in prediction time for the described X link set；

Second acquisition unit, for obtaining the hourage sum in prediction time for the described X link set as described Travelling predicted time between departure place described in prediction time to destination.

Time forecasting methods and device that the present invention provides, due to according to the historical travel between described N bar link link Described N bar link is divided into X link set by time correlation degree, then the path between departure place to destination is examined in processing procedure Consider the degree of correlation on link space, meanwhile, to the corresponding link metric of each link set in described X link set During carrying out ARIMA training, due to using matrix form training it is contemplated that correlation on Link Time so that Differentiated time series smoothness increases, therefore, it is possible to, under improper traffic, improve the standard of predicting travel time Exactness.

Brief description

In order to be illustrated more clearly that the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing Have technology description in required use accompanying drawing be briefly described it should be apparent that, drawings in the following description be only this Some embodiments of invention, for those of ordinary skill in the art, on the premise of not paying creative work, acceptable Other accompanying drawings are obtained according to these accompanying drawings.

Fig. 1 is a kind of time forecasting methods flow chart provided in an embodiment of the present invention；

Fig. 2 is another kind time forecasting methods flow chart provided in an embodiment of the present invention；

Fig. 3 is a kind of path schematic diagram provided in an embodiment of the present invention；

Fig. 4 is a kind of path provided in an embodiment of the present invention division methods flow chart；

Fig. 5 is the acquisition methods flow process of the hourage in prediction time for the 3rd link set provided in an embodiment of the present invention Figure；

Fig. 6 is the historical juncture schematic diagram before prediction time provided in an embodiment of the present invention；

Fig. 7 is a kind of time prediction apparatus structure schematic diagram provided in an embodiment of the present invention.

Specific embodiment

Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete Site preparation description is it is clear that described embodiment is only a part of embodiment of the present invention, rather than whole embodiments.It is based on Embodiment in the present invention, it is every other that those of ordinary skill in the art are obtained under the premise of not making creative work Embodiment, broadly falls into the scope of protection of the invention.

The embodiment of the present invention provides a kind of time forecasting methods, as shown in figure 1, including：

Step 101, the path between departure place to destination is divided into N bar link, described N is more than or equal to 2.

Step 102, according to the historical travel time correlation degree between described N bar link link by described N bar chain k-path partition For X link set, described X is less than or equal to described N.

Step 103, the corresponding link metric of each link set obtaining in described X link set, each described chain Road is gathered corresponding link metric and is made up of the hourage of m historical juncture of all links in described link set, described Historical juncture is the moment in preset time period before prediction time.

Step 104, ARIMA is carried out to the corresponding link metric of each link set in described X link set (Autoregressive Integrated Moving Average Model, difference ARMA model) trains To described X link set prediction time hourage.

Step 105, obtain the hourage sum in prediction time for the described X link set as described prediction time institute State the travelling predicted time between departure place to destination.

So, due to according to the historical travel time correlation degree between described N bar link link by described N bar link It is divided into X link set, then the path between departure place to destination take into account the correlation on link space in processing procedure Degree, meanwhile, during ARIMA training is carried out to the corresponding link metric of each link set in described X link set, Because the training using matrix form is it is contemplated that correlation on Link Time is so that differentiated time series smoothness Increase, therefore, it is possible to, under improper traffic, improve the degree of accuracy of predicting travel time.

Specifically, the embodiment of the present invention provides a kind of time forecasting methods, as shown in Fig. 2 including：

Step 201, the path between departure place to destination is divided into N bar link, described N is more than or equal to 2.

It should be noted that the rule being divided the path between departure place to destination is to arrange as the case may be , for example, by the path between departure place to destination according to principle of equipartition every predeterminable range be divided into one section it is also possible to according to The Actual path planning in path divides, and such as there are 5 crossings between departure place to destination, then will be set out with described 5 crossings Path between ground to destination is mixed and is divided into 6 links, and the present invention does not limit to this.

Step 202, according to the historical travel time correlation degree between described N bar link link by described N bar chain k-path partition For X link set.

Described X is less than or equal to described N.Specifically, the history of often any both links in described N bar link can be obtained The hourage degree of correlation；Then, according in described N bar link often the historical travel time correlation degree of arbitrarily both links generate to Few first link set and/or at least one the second link set, described first link set link order is connected, and The historical travel time correlation degree of arbitrarily both links is more than or equal to predetermined threshold value, the link in described second link set with The historical travel time correlation degree of any one link in described N bar link is both less than described predetermined threshold value；Wherein, generation The number of link set is described X, it should be noted that link order is connected referring to by close departure place to away from departure place Order be connected or by being connected to the order away from destination near destination.

Example it is assumed that the path between departure place to destination is divided into as shown in Figure 34 link by step 201, point Not Wei link L1, L2, L3 and L4, arrow Z indicates direct of travel, then as shown in figure 4, specific link division methods are as follows：

Step 2021, obtain in described link L1, L2, L3 and L4 often the historical travel time correlation of arbitrarily both links Degree.Execution step 2022.

Step 2022, judge that whether the historical travel time correlation degree of link L1 and link L2 is less than predetermined threshold value；If so, Execution step 2023, if it is not, execution step 2024.

Step 2023, generation link set 1 and link set 2, described link L1 are put into described link set 1, by institute State link L2 and put into described link set 2.Execution step 2025.

Step 2024, generation link set 3, described link L1 and link L2 is put into described link set 3.Execution step 2034.

Step 2025, judge that whether the historical travel time correlation degree of link L2 and link L3 is less than predetermined threshold value；If so, Execution step 2026, if it is not, execution step 2027.

Step 2026, generation link set 4, described link L3 is put into described link set 4.Execution step 2031.

Step 2027, described link L3 is put into described link set 2.Execution step 2028.

Step 2028, judge in link L4 and the historical travel time correlation degree of link set 2 link whether there is little Historical travel time correlation degree in predetermined threshold value；If so, execution step 2029, if it is not, execution step 2030.

Step 2029, generation link set 5, described link L4 is put into described link set 5.

Current totally three set, respectively set 1, set 2 and set 5, wherein, have link L1, in set 2 in set 1 There is link L2 and L3, have link L4, set 1 and set 5 to be the second link set in set 5, set 2 is the first link set.

Step 2030, described link L4 is put into described link set 2.

Current totally two set, respectively set 1 and set 2, wherein, have link L1, have link in set 2 in set 1 L2, L3 and L4, set 1 is the second link set, and set 2 is the first link set.

Step 2031, judge that whether the historical travel time correlation degree of link L3 and link L4 is less than predetermined threshold value；If so, Execution step 2032, if it is not, execution step 2033.

Step 2032, generation link set 6, described link L4 is put into described link set 6.

Current totally 4 set, respectively set 1, set 2, set 4 and set 5, wherein, have link L1 in set 1, collection Close and in 2, have link L2, have link L3 in set 4, in set 5, have link L4.Set 1, set 2, set 4 and set 5 are the Two link set.

Step 2033, described link L4 is put into described link set 4.

Current totally 3 set, respectively set 1, set 2 and set 4, wherein, have link L1, have in set 2 in set 1 Link L2, has link L3 and link L4 in set 4.Set 1 and set 2 are the second link set, and set 4 is the first link set Close.

Step 2034, judge in link L3 and the historical travel time correlation degree of link set 3 link whether there is little Historical travel time correlation degree in predetermined threshold value.If so, execution step 2035, if it is not, execution step 2036.

Step 2035, generation link set 6, described link L3 is put into described link set 6.Execution step 2037.

Step 2036, described link L3 is put into described link set 3.Execution step 2040.

Step 2037, judge that whether the historical travel time correlation degree of link L3 and link L4 is less than predetermined threshold value.If so, Execution step 2039, if it is not, execution step 2038.

Step 2038, generation link set 7, described link L4 is put into described link set 7.

Current totally 3 set, respectively set 3, set 6 and set 7, wherein, have link L1, link L2 in set 3, collection Close and in 6, have link L3, in set 7, have link L4.Set 6 is the second link set, and set 3 and set 7 are the first link set.

Step 2039, described link L4 is put into described link set 6.

Current totally 2 set, respectively set 3 and set 6, wherein, have link L1, link L2, in set 6 in set 3 There are link L3, link L4.Set 6 is the second link set, and set 3 is the first link set.

Step 2040, judge in link L4 and the historical travel time correlation degree of link set 3 link whether there is little Historical travel time correlation degree in predetermined threshold value.If so, execution step 2041, if it is not, execution step 2042.

Step 2041, generation link set 8, described link L4 is put into described link set 8.

Current totally 2 set, respectively set 3 and set 8, wherein, have link L1, link L2 and link L3 in set 3, There is link L4 in set 8.Set 8 is the second link set, and set 3 is the first link set.

Step 2042, described link L4 is put into described link set 3.

Current totally 1 set, for set 3, wherein, has link L1, link L2 and link L3 and link L4 in set 3.Collection Closing 3 is the second link set.

Step 203, the corresponding link metric of each link set obtaining in described X link set.

The corresponding link metric of each described link set is by the m historical juncture of all links in described link set Hourage forms, and the described historical juncture is the moment in preset time period before prediction time；The chain being obtained due to step 202 Road set have multiple, the embodiment of the present invention taking the 3rd link set as a example, described 3rd link set be described X link set In any one link set, then the corresponding link metric of each link set obtaining in described X link set includes： Obtain the corresponding link metric of the 3rd link set, the common m row n row of described link metric, described m is institute before described prediction time State the number of the historical juncture in preset time period, often two adjacent described historical junctures be spaced apart predetermined interval, described Prediction time is spaced apart described predetermined interval with the first historical juncture, and described n is the individual of described 3rd link set link Number, described first historical juncture is with described prediction time during immediate history in historical juncture before described prediction time Carve, the number of described m can be arranged as the case may be, and the present invention does not limit to this.

Assume in the 3rd link set totally three articles of links, respectively link L1, link L2 and link L3, then the 3rd link set It is combined into { L1, L2, L3 }, then the corresponding link metric of the 3rd link set is separated by preset time period by with described prediction time The hourage composition of link L1, link L2 and link L3, present invention assumes that described predetermined interval is 5 minutes, described m is 7, then Described preset time period is 5 to 35 minutes, example, as shown in Figure 6 it is assumed that prediction time is 10:00, before prediction time The number of historical juncture is 7, and the historical juncture is respectively 9:55、9:50、9:45、9:40、9:35、9:30、9:25.Then link L1, In link L2 and link L3, the historical juncture corresponding hourage of each link is worth for 7, and link metric H3 obtaining is as follows：

$H3=\left[\begin{array}{ccc}T11& T21& T31\\ T12& T22& T32\\ T13& T23& T33\\ T14& T24& T34\\ T15& T25& T35\\ T16& T26& T36\\ T17& T27& T37\end{array}\right]$

Wherein, T11 to T17 is respectively link L1 in the historical juncture 9:55、9:50、9:45、9:40、9:35、9:30、9:25 Corresponding hourage, T21 to T27 is respectively link L2 in the historical juncture 9:55、9:50、9:45、9:40、9:35、9:30、9: 25 corresponding hourages, T31 to T37 is respectively link L3 in the historical juncture 9:55、9:50、9:45、9:40、9:35、9:30、 9:25 corresponding hourages, the every a line in link metric H3 represents the hourage of same historical juncture difference link, often One row represent the hourage of same link difference historical juncture, it should be noted that in the present embodiment, the travelling of each row Time can be according to the order ascending with the time difference of prediction time, and in practical application, the hourage of each row is permissible Arranging according to the order ascending with the time difference of prediction time can also be according to ascending with the time difference of prediction time Tactic, the present invention does not limit to this.

Described hourage refers to the time spent from origin to destination using default mode of transportation.The present invention Described in default mode of transportation can be for by bus.

Step 204, ARIMA training is carried out to the corresponding link metric of each link set in described X link set Obtain the hourage in prediction time for the described X link set.

Each link set in described X link set is carried out with difference ARMA model ARIMA training Obtain X group training parameter group, described in every group, training parameter group is made up of y training parameter.

So that the 3rd link set is in the acquisition process of the hourage of prediction time as a example, concrete obtaining step such as Fig. 5 institute Show, including：

Step 3011, by described link metric, often in two adjacent row after the data of a line deduct the data of previous row Obtain difference matrix, described difference matrix common m-1 row n row.

For the 3rd link set, the data of a line rear in often adjacent two row in link metric H3 is deducted previous row Data obtains difference matrix M3, that is, the hourage of same link adjacent historical juncture make the difference.

Specifically, difference matrix M3 is as follows：

$M3=\left[\begin{array}{ccc}\mathrm{\Δ}T11& \mathrm{\Δ}T21& \mathrm{\Δ}T31\\ \mathrm{\Δ}T12& \mathrm{\Δ}T22& \mathrm{\Δ}T32\\ \mathrm{\Δ}T13& \mathrm{\Δ}T23& \mathrm{\Δ}T33\\ \mathrm{\Δ}T14& \mathrm{\Δ}T24& \mathrm{\Δ}T34\\ \mathrm{\Δ}T15& \mathrm{\Δ}T25& \mathrm{\Δ}T35\\ \mathrm{\Δ}T16& \mathrm{\Δ}T26& \mathrm{\Δ}T36\end{array}\right]$

Example, Δ T11 is the difference of T12 and T11, and Δ T36 is the difference of T37 and T36, then historical juncture 9:55 is corresponding Difference vector be：

A=[Δ T11 Δ T21 Δ T31]

Historical juncture 9:50 corresponding difference vectors are：

B=[Δ T12 Δ T22 Δ T32]

Historical juncture 9:45 corresponding difference vectors are：

C=[Δ T13 Δ T23 Δ T33]

Historical juncture 9:40 corresponding difference vectors are：

D=[Δ T14 Δ T24 Δ T34]

Historical juncture 9:35 corresponding difference vectors are：

E=[Δ T15 Δ T25 Δ T35]

Historical juncture 9:30 corresponding difference vectors are：

F=[Δ T16 Δ T26 Δ T36]

Step 3012, described difference matrix is divided at least k group differential data so that w group differential data includes institute State the vector of the w of difference matrix to kth+w row.

It should be noted that 1≤w≤k, such as k=3, then the 2nd group of differential data include the 2nd of described difference matrix to The vector of the 5th row, i.e. vectorial B, C, D and E.

Step 3013, respectively by described in every group differential data input ARIMA training obtain described 3rd link set correspond to Training parameter group.

Described ARIMA is：

${\mathrm{\ΔT}}_t=\underset{i=1}{\overset{k}{\Σ}}{\mathrm{\β}}_i{\mathrm{\ΔT}}_{t-i};$

Wherein, described Δ T_tFor the difference vector of the corresponding described difference matrix of historical juncture t, described β_iFor described difference The training parameter of the i-th row of data, 2k is less than or equal to described m-1, the corresponding training parameter group of described 3rd link set by β₁To β_kComposition.

Example it is assumed that k=3, taking the 3rd link set described above as a example, then the 1st to 4 row of described difference matrix Difference vector, the difference vector of the difference vector of the 2nd to 5 row of described difference matrix and the 3rd to 6 row of described difference matrix Difference vector inputs ARIMA：

${\mathrm{\ΔT}}_t=\underset{i=1}{\overset{3}{\Σ}}{\mathrm{\β}}_i{\mathrm{\ΔT}}_{t-i},$

The equation group then obtaining：

D=β₁A+β₂B+β₃C

E=β₁B+β₂C+β₃D

F=β₁C+β₂D+β₃E

Solving equations obtain β₁、β₂And β₃Value, then the corresponding training parameter group of the 3rd link set be β₁、β₂And β₃.

Step 3014, when the corresponding prediction of the 3rd link set is determined according to described training parameter group and described difference matrix The difference vector carved.

Because solving equations obtain β₁、β₂And β₃Value, according to described training parameter group, if the 3rd link set is corresponding The difference vector of prediction time is Z, then bring the value of D, E, F into ARIMA and obtain：

Z=β₁D+β₂E+β₃F；

The value of difference vector Z thus can be obtained.

When step 3015, the difference vector obtaining described 3rd link set corresponding prediction time and described first history Carve corresponding vector sum as the hourage of the 3rd link set of described prediction time.

Difference vector due to the 3rd link set corresponding prediction time is Z, and the first historical juncture, corresponding vector was In step 203, the vector of the first row of matrix H, that is,：

S=[L11 L21 L31]

The hourage of the Z and S sum then obtaining as the 3rd link set of described prediction time, this hourage is Vector, each value in this vector represents the hourage of respective link in prediction time the 3rd link set, that is, in this vector First hourage being worth for prediction time link L1, second hourage being worth for prediction time link L2, the 3rd The individual hourage being worth for prediction time link L3.

The acquisition process of the hourage in prediction time for other link set and the 3rd link in described X link set Set, the present invention repeats no more.

Step 205, obtain described prediction time X link set hourage sum as described prediction time institute State the travelling predicted time between departure place to destination.

In embodiments of the present invention, can be by the hourage sum of links all in all link set of prediction time As described travelling predicted time.

It should be noted that when the current travelling predicted time described in prediction time between departure place to destination of acquisition Afterwards, can be using first historical juncture of this prediction time as next prediction time, execution step 203 obtains new chain again Road matrix, then execution step 204, step 205 obtain described in next prediction time the travelling prediction between departure place to destination Time, particularly obtain departure place described in next prediction time to destination it is also possible to execute from step 201 to step 205 Between travelling predicted time, therefore, time forecasting methods provided in an embodiment of the present invention, when can predict following multiple continuous Carve the travelling predicted time between described departure place to destination.

Time forecasting methods provided in an embodiment of the present invention, due to according to the historical travel between described N bar link link Described N bar link is divided into X link set by time correlation degree, then the path between departure place to destination is examined in processing procedure Consider the degree of correlation on link space, meanwhile, to the corresponding link metric of each link set in described X link set During carrying out ARIMA training, due to using matrix form training it is contemplated that correlation on Link Time so that Differentiated time series smoothness increases, therefore, it is possible to, under improper traffic, improve the standard of predicting travel time Exactness.

The embodiment of the present invention provides a kind of time prediction device 70, as shown in fig. 7, comprises：

First division unit 701, for the path between departure place to destination is divided into N bar link, described N be more than or Equal to 2；

Second division unit 702, for will be described according to the historical travel time correlation degree between described N bar link link N bar link is divided into X link set, and described X is less than or equal to described N；

First acquisition unit 703, for obtaining each link set corresponding link square in described X link set Battle array, the corresponding link metric of each described link set is by the travelling of m historical juncture of all links in described link set Time forms, and the described historical juncture is the moment in preset time period before prediction time；

Difference unit 704, for poor to the corresponding link metric of each link set in described X link set ARMA model ARIMA training is divided to obtain the hourage in prediction time for the described X link set；

Second acquisition unit 705, for obtaining the hourage sum in prediction time for the described X link set as institute State the travelling predicted time between departure place described in prediction time to destination.

So, because the second division unit is according to the historical travel time correlation degree between described N bar link link Described N bar link is divided into X link set, then the path between departure place to destination take into account link in processing procedure The degree of correlation spatially, meanwhile, difference unit enters to the corresponding link metric of each link set in described X link set During row ARIMA training, because the training using matrix form is it is contemplated that correlation on Link Time is so that poor Time series smoothness after point increases, therefore, it is possible to, under improper traffic, improve the accurate of predicting travel time Degree.

Further, described time prediction weighing device can be computer, and described N article of chain route the 1st is to N link group Become, the second division unit 702 specifically for：

Obtain the historical travel time correlation degree of often any both links in described N bar link；

According in described N bar link often the historical travel time correlation degree of arbitrarily both links generate at least one first chain Road set and/or at least one second link set, described first link set link order is connected, and any both links Historical travel time correlation degree be more than or equal to predetermined threshold value, the link in described second link set and described N bar link In any one link historical travel time correlation degree be both less than be set forth in predetermined threshold value；

Wherein, the number of the link set of generation is described X.

Described first acquisition unit 703 specifically for：

Obtain the corresponding link metric of the 3rd link set, described 3rd link set is in described X link set Any one link set, described link metric common m row n row, described m is in described prediction time as described before preset time period Historical juncture number, often two adjacent described historical junctures be spaced apart predetermined interval, described prediction time and first Historical juncture be spaced apart described predetermined interval, described n is the number of described 3rd link set link, described first history Moment be in historical juncture before described prediction time with described prediction time the immediate historical juncture.

Described difference unit 704 specifically for：

By in corresponding for described 3rd link set link metric, often in two adjacent row after a line data deduct previous The data of row obtains difference matrix, and the common m-1 row n of described difference matrix arranges；

Described difference matrix is divided at least k group differential data so that w group differential data includes described difference matrix W to kth+w+1 row vector, 1≤w≤k；

Respectively differential data input ARIMA training described in every group is obtained the corresponding training parameter of described 3rd link set Group, described ARIMA is：

${\mathrm{\ΔT}}_t=\underset{i=1}{\overset{k}{\Σ}}{\mathrm{\β}}_i{\mathrm{\ΔT}}_{t-i};$

Wherein, described Δ T_tFor the difference vector of the corresponding described difference matrix of historical juncture t, described β_iFor described difference The training parameter of the i-th row of data, 2k is less than or equal to described m-1, the corresponding training parameter group of described 3rd link set by β₁To β_kComposition；

Determine the difference of the 3rd link set corresponding prediction time according to described training parameter group and described difference matrix Vector；

The difference vector obtaining described 3rd link set corresponding prediction time is corresponding with described first historical juncture Vectorial sum is as the hourage of the 3rd link set of described prediction time.

Described predetermined interval is 5 minutes, and described k is 6.

Time prediction device provided in an embodiment of the present invention, because the second division unit is according to described N bar link link Between historical travel time correlation degree described N bar link is divided into X link set, then the path between departure place to destination Take into account the degree of correlation on link space in processing procedure, meanwhile, difference unit is to each chain in described X link set Road gathers during corresponding link metric carries out ARIMA training, because the training using matrix form is it is contemplated that link Temporal correlation so that differentiated time series smoothness increases, therefore, it is possible under improper traffic, Improve the degree of accuracy of predicting travel time.

One of ordinary skill in the art will appreciate that：The all or part of step realizing said method embodiment can be passed through Completing, aforesaid program can be stored in a computer read/write memory medium the related hardware of programmed instruction, this program Upon execution, execute the step including said method embodiment；And aforesaid storage medium includes：ROM, RAM, magnetic disc or light Disk etc. is various can be with the medium of store program codes.

The above, the only specific embodiment of the present invention, but protection scope of the present invention is not limited thereto, and any Those familiar with the art the invention discloses technical scope in, change or replacement can be readily occurred in, all should contain Cover within protection scope of the present invention.Therefore, protection scope of the present invention should be defined by described scope of the claims.

Claims (10)

1. a kind of time forecasting methods are it is characterised in that include：

Path between departure place to destination is divided into N bar link, described N is more than or equal to 2；

Described N bar link is divided into by X link set according to the historical travel time correlation degree between described N bar link link Close, described X is less than or equal to described N；

Obtain the corresponding link metric of each link set in described X link set, each described link set is corresponding Link metric was made up of the hourage of m historical juncture of all links in described link set, and the described historical juncture is pre- Moment in preset time period before the survey moment；

Difference ARMA model is carried out to the corresponding link metric of each link set in described X link set ARIMA training obtains the hourage in prediction time for the described X link set；

Obtain the hourage sum in prediction time for the described X link set as departure place described in described prediction time to mesh Ground between travelling predicted time.

2. method according to claim 1 is it is characterised in that described N article of chain route the 1st is to N link composition,

Described described N bar link is divided into by X link according to the historical travel time correlation degree between described N bar link link Set includes：

Obtain the historical travel time correlation degree of often any both links in described N bar link；

According in described N bar link often the historical travel time correlation degree of arbitrarily both links generate at least one first link set Close and/or at least one second link set, described first link set link order is connected, and the going through of any both links The history hourage degree of correlation is more than or equal to predetermined threshold value, in the link in described second link set and described N bar link The historical travel time correlation degree of any one link is both less than described predetermined threshold value；

Wherein, the number of the link set of generation is described X.

3. method according to claim 1 it is characterised in that

The described corresponding link metric of each link set obtaining in described X link set includes：

Obtain the corresponding link metric of the 3rd link set, described 3rd link set is any in described X link set One link set, described link metric common m row n row, described m is going through in described prediction time as described before preset time period The number in history moment, often two adjacent described historical junctures be spaced apart predetermined interval, described prediction time and the first history Moment be spaced apart described predetermined interval, described n is the number of described 3rd link set link, described first historical juncture For in the historical juncture before described prediction time with described prediction time the immediate historical juncture.

4. method according to claim 3 it is characterised in that

Described difference auto regressive moving average is carried out to the corresponding link metric of each link set in described X link set Model ARIMA training obtains described X link set and includes in the hourage of prediction time：

By in corresponding for described 3rd link set link metric, often in two adjacent row, the data of rear a line deducts previous row Data obtains difference matrix, described difference matrix common m-1 row n row；

Described difference matrix is divided at least k group differential data so that w group differential data includes the of described difference matrix W to kth+w+1 row vector, 1≤w≤k；

Respectively differential data input ARIMA training described in every group is obtained the corresponding training parameter group of described 3rd link set, Described ARIMA is：

${\mathrm{\ΔT}}_t=\underset{i=1}{\overset{k}{\Σ}}{\mathrm{\β}}_i{\mathrm{\ΔT}}_{t-i};$

Wherein, described Δ T_tFor the difference vector of the corresponding described difference matrix of historical juncture t, described β_iFor described differential data The i-th row training parameter, 2k is less than or equal to described m-1, and the corresponding training parameter group of described 3rd link set is by β₁To β_k Composition；

Determine the difference vector of the 3rd link set corresponding prediction time according to described training parameter group and described difference matrix；

Obtain the difference vector vector corresponding with described first historical juncture of described 3rd link set corresponding prediction time Sum is as the hourage of the 3rd link set of described prediction time.

5. method according to claim 4 it is characterised in that

Described predetermined interval is 5 minutes, and described k is 6.

6. a kind of time prediction device is it is characterised in that include：

First division unit, for the path between departure place to destination is divided into N bar link, described N is more than or equal to 2；

Second division unit, for according to the historical travel time correlation degree between described N bar link link by described N bar link It is divided into X link set, described X is less than or equal to described N；

First acquisition unit, for obtaining the corresponding link metric of each link set in described X link set, each institute State the corresponding link metric of link set to be made up of the hourage of m historical juncture of all links in described link set, The described historical juncture is the moment in preset time period before prediction time；

Difference unit, returns certainly for the corresponding link metric of each link set in described X link set is carried out with difference Moving average model(MA model) ARIMA training is returned to obtain the hourage in prediction time for the described X link set；

Second acquisition unit, for obtaining the hourage sum in prediction time for the described X link set as described prediction Travelling predicted time between departure place described in the moment to destination.

7. time prediction device according to claim 6 is it is characterised in that described N article of chain route the 1st is to N link group Become,

Described second division unit, specifically for：

Obtain the historical travel time correlation degree of often any both links in described N bar link；

According in described N bar link often the historical travel time correlation degree of arbitrarily both links generate at least one first link set Close and/or at least one second link set, described first link set link order is connected, and the going through of any both links The history hourage degree of correlation is more than or equal to predetermined threshold value, in the link in described second link set and described N bar link The historical travel time correlation degree of any one link is both less than described predetermined threshold value；

Wherein, the number of the link set of generation is described X.

8. time prediction device according to claim 6 it is characterised in that

Described first acquisition unit specifically for：

Obtain the corresponding link metric of the 3rd link set, described 3rd link set is any in described X link set One link set, described link metric common m row n row, described m is going through in described prediction time as described before preset time period The number in history moment, often two adjacent described historical junctures be spaced apart predetermined interval, described prediction time and the first history Moment be spaced apart described predetermined interval, described n is the number of described 3rd link set link, described first historical juncture For in the historical juncture before described prediction time with described prediction time the immediate historical juncture.

9. time prediction device according to claim 8 it is characterised in that

Described difference unit specifically for：

By in corresponding for described 3rd link set link metric, often in two adjacent row, the data of rear a line deducts previous row Data obtains difference matrix, described difference matrix common m-1 row n row；

Described difference matrix is divided at least k group differential data so that w group differential data includes the of described difference matrix W to kth+w+1 row vector, 1≤w≤k；

Respectively differential data input ARIMA training described in every group is obtained the corresponding training parameter group of described 3rd link set, Described ARIMA is：

${\mathrm{\ΔT}}_t=\underset{i=1}{\overset{k}{\Σ}}{\mathrm{\β}}_i{\mathrm{\ΔT}}_{t-i};$

Wherein, described Δ T_tFor the difference vector of the corresponding described difference matrix of historical juncture t, described β_iFor described differential data The i-th row training parameter, 2k is less than or equal to described m-1, and the corresponding training parameter group of described 3rd link set is by β₁To β_k Composition；

Determine the difference vector of the 3rd link set corresponding prediction time according to described training parameter group and described difference matrix；

Obtain the difference vector vector corresponding with described first historical juncture of described 3rd link set corresponding prediction time Sum is as the hourage of the 3rd link set of described prediction time.

10. time prediction device according to claim 9 it is characterised in that

Described predetermined interval is 5 minutes, and described k is 6.